Author

Topic: If you can explain this, I will send 1 BTC your way... (Read 2710 times)

copper member
Activity: 56
Merit: 0
I like the resistance idea.

First and second card power up, cause a drain when the driver inits them and poof....

Same scenario with a different cable and it's microvolts away from malfunctioning but somehow survives.
legendary
Activity: 1708
Merit: 1010
Cord goes in, screen goes out

Can't explain that

Never a misscommunication

Hidden dead short across a pair of pins?
hero member
Activity: 695
Merit: 502
PGP: 6EBEBCE1E0507C38
Cord goes in, screen goes out

Can't explain that

Never a misscommunication
full member
Activity: 518
Merit: 100
The answer of resitance is the villian. On a standard setup, you can loose about 10k per card with an entension cable due to the increased time it takes to send a signal. It may also prevent a bootup.

Then why would it work with another cable, arguably exactly the same materials/length/makeup/manufacturing processes... etc etc..?

Probably a combination of a faulty cable and a slot that delivers less stable power than the other slots, which could be able to compensate for the problem.
JJG
member
Activity: 70
Merit: 20
=Why on earth would an extension cable stop only the desktop from booting when used in card slot number 1?

Not sure if this is your issue or not, but if by saying card slot #1 you actually mean the first card slot (number zero) then the bios might be misidentifying the card as a bootable network card, waiting for it to initialize.

Keeping the same slot and switching the one "cable from hell" changes my results, so I doubt this is the case.

High speed signals require great care.

Proper transmission requires traces with the proper impedance and a minimum of discontinuities in the signal path. Each discontinuity (connectors, pins, etc) will pass most of the signal, but a small portion will be reflected backward toward the source. The bigger the discontinuity, the larger the reflection.

Normally, your traces are relatively short on the motherboard and these reflections will reach the source before the next symbol (think 'bit' in this case) is transmitted. When you extend the trace length as far as you have, it is entirely possible that the reflections take so long to reach the source that they're overlapping on the next symbol (bit). This is called inter-symbol interference (ISI).

So why does it only happen in slot 1? My guess is that this slot has either the longest traces, or a slightly different impedance that causes a larger discontinuity when you connect your extender contraption.

The solution is to not double up your extenders. If you must extend something that far, use a single extender.

Edit: Note that when I say impedance, I'm not referring to just resistance. The complex impedance of a transmission line or lumped element refers to how it interacts with a time-varying signal. It's not the same quantity that you can measure with your multimeter.
sr. member
Activity: 280
Merit: 252
The answer of resitance is the villian. On a standard setup, you can loose about 10k per card with an entension cable due to the increased time it takes to send a signal. It may also prevent a bootup.

Then why would it work with another cable, arguably exactly the same materials/length/makeup/manufacturing processes... etc etc..?
sr. member
Activity: 280
Merit: 252
Interference from using a card longer than PCI spec allows for in that first slot. Just the magic combination of that specific cable in that specific slot.

I will test this theory shortly...

You are right.. the cables are much too close together.. they are quite thick and are therefore tough to maneuver... so they all generally lay on top of eachother.

However when I use the same cable as the "upper" portion of the double connection... it rests higher up on my case and does not come into contact or even come close to any other cables.. so I am not sure this is it either.
sr. member
Activity: 280
Merit: 252
=Why on earth would an extension cable stop only the desktop from booting when used in card slot number 1?

Not sure if this is your issue or not, but if by saying card slot #1 you actually mean the first card slot (number zero) then the bios might be misidentifying the card as a bootable network card, waiting for it to initialize.

Keeping the same slot and switching the one "cable from hell" changes my results, so I doubt this is the case.
sr. member
Activity: 280
Merit: 252
I really have no idea if this is your issue but under linux whenever I reconfigure the locations of my cards or add new ones, at the point where it hangs, I can SSH in and rerun the original ati configure command sudo aticonfig --initial -f --adapter=all -- allow it to replace the xorg.conf and then reboot. Then things start working again.

I had thought of that too, so I re-created my linuxcoin bootable usb from scratch and made sure persistence was not enabled at any time. Linuxcoin is very good at recognizing new cards no matter what their positions.

Anyways the positions of the cards and/or the brands do not matter, the only "problem" is using this one specific "cable from hell" i have labeled it now...  that does not work as an extension but does work as a single cable. And not only that... it DOES work as an extension, it simply does not show the desktop.

My only idea is that perhaps since the cables are very inexpensive... if the desktop requires more juice to flow through the cable number 1 (in order to display it..?) then that seems to be the only reason in my mind that this could be happening.

So confused... but thankfully after identifying that problem cable I now have 6 cards up and running on my MSI 790FXA... 3 different brands. (I was worried it was the interchanging of brands at first as well, but obviously not..).
full member
Activity: 168
Merit: 100
A couple years ago, I experimented with a PCI VGA card that you could extend the signal up to 30ft through the use of a Ethernet cable.

The interesting thing was in some cases if the cable was too long, like 50ft or so, this would cause some elements of the windows desktop to not load properly.  (Desktop background, icons etc.)

This also caused CPU usage to spike significantly.

In short what was happening is the timing signal for the PCI device was off, and this caused the PCI VGA driver to freak out, causing excessive CPU usage and causing some graphical elements to fail to load.

I believe this is exactly what is happening in your case.  Too much resistance in the PCI-E slot, screws up the timing signal and freaks out the GPU driver.
sr. member
Activity: 406
Merit: 250
QUIFAS EXCHANGE
The answer of resitance is the villian. On a standard setup, you can loose about 10k per card with an entension cable due to the increased time it takes to send a signal. It may also prevent a bootup.
newbie
Activity: 22
Merit: 0
My anwser is resistance. Maybe after loading OS voltage is little dropping, so that speciffic cable creates too big resistance. (I think, that after loading OS card turns idle - less power used - too big resistance - motherboard thinks there is no card). It would be more doubtful if other cards turns idle too (in other slots).

Other slots could not have power drop added (as power saving or sth like that)


resistance is futile
member
Activity: 98
Merit: 10
My anwser is resistance. Maybe after loading OS voltage is little dropping, so that speciffic cable creates too big resistance. (I think, that after loading OS card turns idle - less power used - too big resistance - motherboard thinks there is no card). It would be more doubtful if other cards turns idle too (in other slots).

Other slots could not have power drop added (as power saving or sth like that)
inh
full member
Activity: 155
Merit: 100
Interference from using a card longer than PCI spec allows for in that first slot. Just the magic combination of that specific cable in that specific slot.
legendary
Activity: 1708
Merit: 1010
=Why on earth would an extension cable stop only the desktop from booting when used in card slot number 1?

Not sure if this is your issue or not, but if by saying card slot #1 you actually mean the first card slot (number zero) then the bios might be misidentifying the card as a bootable network card, waiting for it to initialize.
sr. member
Activity: 280
Merit: 252
Prepare to have your mind blown. I know mine is.

I purchased a bunch of pcie 1x extension cables through ebay. Modded them myself, and shorted connections A1+B17 with some solid wiring.

I occasionally combine two cables together to double their lengths. This works pretty well in general.

I have been wracking my brain for days trying to figure out why certain combos of cards didn't work in my MSI 890FXA-GD70 mobo.

It turns out that a specific cable... when used as a "doubling" cable on card number 0/1 (first card)... will show MOST of the booting sequence in linuxcoin v0.2a...

I mean I can see the post screen, the boot sequence of black and white screen/text... and then when loading the desktop everything hangs...

If I change that one cable out for another, the desktop loads fine.

If I use that cable in the 2nd - Nth card, the desktop loads fine.

Why on earth would an extension cable stop only the desktop from booting when used in card slot number 1?

Answer me this and you get 1 BTC.

Extra info:
- I have tried xfx, diamond and sapphire cards with the same cable... same result.
Jump to: